Biosensors and Bioelectronics 90 (2017) 298307

Contents lists available at ScienceDirect

Biosensors and Bioelectronics

journal homepage: www.elsevier.com/locate/bios

Increasing trend of wearables and multimodal interface for human activity MARK
monitoring: A review

Preeti Kumari , Lini Mathew, Poonam Syal
Department of Electrical Engineering, National Institute of Technical Teacher's Training and Research, Sector-26, Chandigarh 160019, India

A R T I C L E I N F O A BS T RAC T

Keywords: Activity recognition technology is one of the most important technologies for life-logging and for the care of
Human activity monitoring elderly persons. Elderly people prefer to live in their own houses, within their own locality. If, they are capable to
Human Computer Interface do so, several benets can follow in terms of society and economy. However, living alone may have high risks.
Wearable sensors Wearable sensors have been developed to overcome these risks and these sensors are supposed to be ready for
Smart sensors
medical uses. It can help in monitoring the wellness of elderly persons living alone by unobtrusively monitoring
Multimodal interface
Biomedical
their daily activities. The study aims to review the increasing trends of wearable devices and need of multimodal
Shared control architecture recognition for continuous or discontinuous monitoring of human activity, biological signals such as
Electroencephalogram (EEG), Electrooculogram (EOG), Electromyogram (EMG), Electrocardiogram (ECG)
and parameters along with other symptoms. This can provide necessary assistance in times of ominous need,
which is crucial for the advancement of disease-diagnosis and treatment. Shared control architecture with
multimodal interface can be used for application in more complex environment where more number of
commands is to be used to control with better results in terms of controlling.

1. Introduction
Activity monitoring aims to monitor the actions of agents obtained
from a number of observations on the actions of agents and conditions
of the environment. Activity recognition plays an important role in
ambient living environments to assess changes from the normal
behavior of elderly people (Uslu et al., 2013). The objective of activity
monitoring is to analyze or interpret the ongoing events from data
automatically. Since 1980s, this eld has grasped the attention of many
researchers due to its ability to provide personalized support for several
applications which include patient monitoring, surveillance and many
dierent varieties of systems involving interactions between machines
and persons such as Brain-Computer Interfaces (BCI).
Activity monitoring includes two processes: data acquisition fol-
lowed by classication of acquired data. The acquisition of data
includes acquiring the bio-signals and signal preprocessing. The bio-
signals can be EEG, EOG, EMG or ECG depending upon the applica-
tion. Signal preprocessing includes amplifying, ltering, averaging,
extracting relevant features to be used as training data for classier etc.
For classication, various methods are used such as least squares
(Marquardt, 1963), Knearest neighbors (k-NN) (Cunningham and
Delany, 2007), hidden Markov method, articial neural networks
(ANN) (Hopeld, 1988) decision tree classication and support vector
machines (SVM) (Gunn, 1998). Data acquisition process has two
dierent approaches: one is the traditional approach which uses
external sensors such as cameras or other monitoring devices (Lin,
2009) and the second one is the newly introduced approach which uses
wearable wireless sensors. Both the approaches use dierent types of
electrodes to acquire the physiological signals. These electrodes can be
active or passive electrodes as per requirement and the positions of the
electrode placement can be signal-dependent as well as application-
dependent. For example, to acquire the EOG signal for vertical move-
ment of eyeballs, electrodes are placed above and below the eyes and
for horizontal movement of eyeballs, electrodes are placed on right side
of right eye and left side of left eye. For EEG recording, electrodes are
placed according to the 1020 International system. In the traditional
approach, sensors are xed at predetermined places, so the conjectures
are fully based on the discretionary interaction between person under
monitoring and sensors used. Examples of external sensing approach
are Intelligent Homes (Englebienne and Krose, 2010; Tolstikov et al.,
2011; Yang et al., 2011; Sarkar et al., 2011). However, in the second
approach, sensors are attached to the human body. Human Activity
Recognition (HAR) systems based on wearable sensors can be categor-
ized in two stages. One is learning approach, may be supervised,
unsupervised or semi-supervised. In the second stage, depending on
the response time, these approaches may be either oine or online.

Corresponding author.
E-mail addresses: pgrr.2403@gmail.com (P. Kumari), lenimathew@yahoo.com (L. Mathew).

http://dx.doi.org/10.1016/j.bios.2016.12.001
Received 15 September 2016; Received in revised form 1 December 2016; Accepted 1 December 2016
Available online 02 December 2016
0956-5663/ 2016 Elsevier B.V. All rights reserved.
P. Kumari et al.
Fig. 1. Approaches of Human Activity Monitoring (Lara and Labrador, 2013).
Online approaches immediately recognize the action performed and
give feedback accordingly. While oine schemes require more time to
respond to the actions performed. Oine scheme demands high
computation and is suitable for applications that do not demand
immediate feedback in real-time. The hierarchy of these two ap-
proaches has been shown in Fig. 1. These two approaches of HAR
systems are used for dierent purposes and both are having dierent
challenges (Lara and Labrador, 2013). Continuous increase in popula-
tion besides a consequent aging portion has forced rapid rises in
health-care and health maintenance costs. Apart from that, there are
many technical challenges at the designing point of view of activity
monitoring systems. As even the same person does not act in the same
way for the same reason every time and some dierent activities may
show similar behavior. Due to these uncertainties and randomness,
recognition accuracy decreases signicantly. To overcome this pro-
blem, the healthcare or HAR system is searching for some systems
having wearable wireless sensors in which continuous monitoring of
patients is possible in real time even without hospitalization. This may
be a complete transformation of existing healthcare system. There are
various applications of real-time activity monitoring systems. In
clinical applications, continuous monitoring of physically or mentally
disabled inhabitants has become important for their safety. Similarly,
interactive or virtual games like simulators may improve person's
experience and can produce more enjoyable game virtually.
The block diagram of a simplest architecture of the HAR system is
represented in Fig. 2. A number of sensors are used to handle dierent
monitoring tasks (Malhi, 2010). Sensors may be used to measure
attributes such as motion, location, temperature, ECG (Iglesias et al.,
2011; Choujaa and Dulay, 2008; Juha et al., 2006; Jatob et al., 2008).
Fig. 2. Representation of Human Activity Recognition (HAR) system in block diagram (Malhi, 2010).
299
Biosensors and Bioelectronics 90 (2017) 298307
Sensor data is collected and after analysis of acquired data, made these
data available to the patients, caretakers, wearers or healthcare
professionals. The goal is to improve the management and care-
delivery, to engage patients and encourage independent living
(Rodgers et al., 2015).
2. Wearable Sensors
Wearable sensors are typically wireless tiny sensors enclosed in
bandages or some patches or something that can be worn. It may be a
ring, shirt, skin patches, watch, or exoskeletons. Ring sensor and smart
shirt are shown in Fig. 3(a) and Fig. 3(b). Sensitive biological elements,
transducer, and associated electronics are the components of wearable
biosensors. Calorimetric, Potentiometric, Amperometric, Optical,
Piezo-electric biosensors and immunosensors are dierent types of
wearable sensors. The data acquired from these wearable sensors are
processed as per requirement for a particular application. Wearable
sensors are completely unobtrusive devices that help physicians to
overwhelm the restrictions of traditional technologies. Through wear-
able systems, biological signals can be continuously acquired wirelessly
and thus patients can be monitored remotely. These sensors have
applications for the persons suering from severe diseases like
Parkinson disease or heart-attack (Mariani et al., 2013; Chen et al.,
2011). For example, ring sensor shown in Fig. 3(a) is used to monitor
heart rate and oxygen saturation. It is an optical biosensor based pulse
oximetry sensor. Every time when the heart muscle contracts, a
pressure pulse is passed through circulatory system. This pressure
pulse causes displacement in vessel walls when the pulse travels
through the vessels. This displacement changes by photoelectric
method and can be measured at dierent points on the human body
to detect pulsatile blood volume. Light is emitted from LED placed on
the wearable ring sensor shown in Fig. 3(a) and travelled through
artery. Blood is forced to extreme points with every heart contraction
and blood ow amount increases in the nger. As a result, optical
density of transmitted light through the nger reduces. Therefore,
functioning of heart is monitored just by measuring alteration in this
optical density. Ring sensor also has a trans-receiver for bidirectional
communication and for uploading the data at a point. However, ring
sensor has a major disadvantage that a limited number of physiological
parameters can be monitored with this sensor.
A wearable smart shirt shown in Fig. 3(b) is a device which tracks
some vital analytes of human body such as breathing rate, body
temperature, respiration rate etc. by using optical and electrical bers.
These bers are conductive and are woven into the fabric of the shirt.
The biological sensors are used to sense the presence as well as
concentration of a substance to be monitored. It can be very useful
for sports performance monitoring, medical care, hazardous applica-
P. Kumari et al.
Fig. 3. (a). Ring Sensor & its Schematic Diagram. (b). Smart Shirt and its Sensory Architecture.
tions, reman cloths, diving cloths or cloths for any specic function we
want.
Other applications in health care are biofeedback respiratory
system (Liu et al., 2011), mental stress assessment (Seoane et al.,
2014), rehabilitation system (Patel et al., 2012), weight lifting exercise
(Velloso et al., 2013), vision sensor in real-time (Peixoto et al., 2001),
posture & movement wearable sensor (Ugulino et al., 2012). Apart
from health-care systems, wearables are widely used in various other
elds such as to provide training support for athletes, to monitor the
persons who work at hazardous places. It also facilitates physically-
disabled people to continue enjoying their independence for longer
time span. It can give mental peace to the elderly persons through a
simple panic button (Versel, 2012; Alert Devices for the Elderly, 2016;
Mobile phones to come with panic button from, 2017, 2016) and
provides assurance to family members and friends.
2.1. Basic requirements of a wearable device
Before developing a wearable system, it is essential to have a clear
idea about the basic requirements and designing challenges for any
wearable device. There are always hardware and software constraints
beginning from low-energy operations, light-weight and safety require-
ments. While person is placing the wearable sensor on his/her body,
the chances of thermal injury must be considered and should be
reduced by controlling the sensing and wireless frequency and radio
duty cycle of wearable sensor. A novel optimization framework has
been proposed for considering safety requirements as well as sustain-
ability requirements based on the human physiology and system level
300
Biosensors and Bioelectronics 90 (2017) 298307
design parameters have been derived (Bagade et al., 2014). Some basic
requirements are discussed below:
2.1.1. Aesthetics
Appearance is so important that many top companies are working
in partnership with the fashion industry to make these wearable
devices fashionable and more attractive.
2.1.2. Size
Wearable devices must be compact enough so that they can easily t
onto the human body. It should be comfortable to the wearer as well as
they must have more features integrated into the same space.
2.1.3. Water tolerance
As the wearable device has to go everywhere along with human
body, it should be tolerant of the environmental conditions such as
temperature, water droplets, moisture, and sweat for continuous
monitoring of the human behavior.
2.1.4. Power consumption
Wearable devices are powered by a battery. For monitoring of
human behavior continuously 247, power requirement will be very
high. Among all the system-components, communication system con-
sumes most of the energy. An autonomous wireless wearable sensor
node having exible energy harvesting mechanism has been presented
which is equipped with ultralow power management circuit. This
exible power management circuit can transfer maximum electrical
power from solar energy to give power to the wireless sensor node (Toh
P. Kumari et al.
et al., 2014). Reduction of power consumption of these devices poses
requirement of special features in Micro Controller Unit and rmware
algorithms. ARM architecture is very popular for wearable systems
because of its best performance. Also, ANT+, Bluetooth Low Energy
(BLE) are energy ecient wireless technologies.
2.1.5. Wireless communication
For continuous monitoring of human behavior, wireless connectiv-
ity is essential, as the person or the wearer cannot move freely so far
with wires. Also, the person will not feel comfortable being surrounded
by wires. Apart from this, they may need to interact with more than one
device. According to the type and features oered, the device may
support dierent wireless protocols such as Wi-Fi, ANT+, and IEEE
802.15.4 based protocol. Dierent data source may generate heavy
time-varying trac which may lead to intolerant abeyancy in wireless
wearable sensors. Moreover, delivery of data in real-time, packet loss
and fading in data transmission induced by movement as well as
surrounding environment is a big challenge.
2.1.6. Operating system
Sometimes, the wearable device may need a specic operating
system such as Android depending on the features being oered. For
example, a smart watch designed to be an extension of a mobile phone
needs to run a specic advanced operating system.
Day by day, new trends can be seen in the eld of Wearable system
which has enhanced features. For example, shirt or other cloths with
all-fabric keyboard made by conductive thread can be washed in the
machine same as ordinary cloths. So, it is water durable which is one of
the basic requirements for a wearable device. Computerized cloths can
be the next generation for computers and other devices which does not
require strap of electronics into our body. Although a huge amount of
eort is being made in the wearable sensors, challenges like user-
acceptance, low power consumption, interference in wireless systems
are still to be resolved for better usability and functionality of these
wearable devices.
Fig. 4. Block diagram representation of a wireless wearable system (Ramasamy et al., 2014).
301
Biosensors and Bioelectronics 90 (2017) 298307
3. Architecture of wearable system
Architecture of wearable system can be explained with the help of a
block diagram representation as shown in Fig. 4. It consists of dierent
blocks namely: Power supply, display, wireless connectivity, motion
sensing and application processor. Micro-Electro-Mechanical Systems
(MEMS) sensors monitor the movement of a human body in every
dimension. The examples of motion sensor or MEMS sensor may
include accelerometers, magnetometers, and gyroscopes. Analogue
sensors are biometric sensors such as heart rate monitors and EEG,
EMG. Analogue Front End (AFE) basically preprocessed the data
acquired from sensors. It contains operational ampliers, lters, and
A/D Convertors. User Interface (UI) systems are used for interaction
between human being and wearable device which is an important
consideration.
Capacitive touch sensing is the most preferred User Interface
available today. UI elements are indicators used for alerts implementa-
tion from the device to the user as well as from user to the device. Pulse
width modulation (PWM) is used to drive these UI elements. Among all
these components of wearable system, one of the most important
components is application processor and is selected very carefully. The
selection of application processor depends on the type, features and
characteristics of the device. The latest microcontrollers are compatible
with most of the wearable systems. MCUs are compact and can
integrate a number of functions on a single chip which is very
important to reduce the overall size and cost of a wearable device.
For example, 32 bit ARM architecture is a very popular Central
processing Unit (CPU) technology for wearable systems as it shows
best performance in terms of computation and energy eciency
(Ramasamy et al., 2014). For Some advanced applications, wearable
systems may need a separate co-processor to ooad the sensor data
from the main processor. This is needed when the system has to load
the sensor data and at the same time, it has to analyze it together in real
time, requiring full CPU attention. This function is known as sensor
hub or sensor fusion.
P. Kumari et al.
4. Current market situation and future trend of wearable
devices
Wearable technology broadly concerns electronic devices, apparel
and textiles. The Fig. 5 shows two types of wearable technology and its
characteristics developed by many big companies. The wearable device
business has been already grown up. IDTechEx has been already grown
up. IDTechEx has examined and explained leading signs of future
wearable technology sales like Google Trends, reduced cost of these
technologies, continuous increase in most promising features, user-
friendliness and starting sales of newly launched smart wrist-wears (for
example Samsung watch, tness monitors). All of these show that very
fast growth is in prospect.
There are a number of big names namely Google, Apple, Adidas,
Nike, Samsung, and Intel behind the most advanced and promising
new developments in the eld of wearable technology (Hayward et al.,
2015). The market for wearable technology is entering into a rapid
growth phase. However, the healthcare sector will remain dominant
which includes wellness and medical tness. At the end of the next
decade, advanced technologies such as wearable electronics will be
matched with the healthcare market, having new healthcare-infor-
matics devices challenging sales potential in billion dollars.
The pie chart in Fig. 6 shows that North America is more forward
towards wearable electronics developers and manufacturers for med-
ical purpose, tness and wellness as compared to other territories like
Europe, East Asia.
A report provided by IDTechEx (Hayward and Chansin, 2016)
shows detailed description about sensor types which is a dominating
component of any wearable technology products. The relative market
size by wearable sensor types in 2020 is shown in Fig. 7. With the most
Fig. 5. The two main types of wearable technology and their characteristics (Hayward
et al., 2015).
Europe
North America
East Asia
others
Fig. 6. A comparison of dierent territories in the eld of development and manufacture
of wearable electronics for healthcare (Hayward et al., 2015).
302
Biosensors and Bioelectronics 90 (2017) 298307
Fig. 7. The Relative Market Size by wearable sensor types in 2020 (Hayward and
Chansin, 2016).
featured and promising sensor options, the report concluded that there
will be approximately 3 billion wearable sensors by 2025 and most of
them will be newly emerged.
5. Recent developments in human activity monitoring using
wearable sensors and bio-potential signals
Wearable sensors have vast scope in dierent areas such as medical
services, surveillance training support for athletes, automation indus-
try, connected smart homes, robot control, and security systems.
However, at designing end, there are some challenges including user
acceptance, small and comfortable to wear (wearability), hardware
design, development, and support, prototyping, manufacturing, and
deployment, real-time data collection and processing (reliable commu-
nication), large-scale user studies (interoperability), cost eective,
energy consumption, privacy and security problems. To clear these
hurdles, many research works are going on. Few of them are discussed
here. Considering the power/energy consumption requirements of
wireless sensor networks, Patel et al. (2009) have developed a wearable
system integrated with some algorithms to diagnose the severity of
diseases and motor diculties in patients. This proposed conguration
allows monitoring patients for many days, once the batteries of the
Body Sensor Network (BSN) nodes are fully charged. P. Bonato has
talked about the new trends in combining wearable technology and
robots (Bonato, 2010). Integrating robots and wearable technology
seem like a key step in the direction of achieving the goal of monitoring
patients at home, eectively. These systems are complex and are able to
monitor the status of the subjects and provide a tool which is invaluable
to tackle the emergency situations. A wearable ECG signal sensing and
processing system have been proposed. The wearable sensor is
combined with a wireless protocol (ANT protocol) for data transfer.
To reduce the power requirement and eective sensor-size, the wireless
protocol was used as a low data rate module (Nemati et al., 2012). The
proposed system can be xed on a T-shirt and the size of the electrodes
was smaller than many recent works. A wearable module and articial
neural network (ANN) based activity/ behavior classication algorithm
has been presented by (Lin et al., 2012) to estimate energy expenditure.
The purpose of the design was to classify human body activities with
similar intensity levels and then to develop energy expenditure regres-
sion models using ANN for optimization of the estimation perfor-
mance. The categorization of human body activities use the acceler-
ometer data and ECG signal data acquired by wearable sensor modules.
A human emotion recognition system is shown in Fig. 8.
Quazi (2012) has reported in his M.S. thesis that the data from the
wearable devices may also be used to determine dierent emotions of
the person under monitoring. The system was based on information
provided by the physiological signals obtained from a skin-temperature
and skin-conductance sensor. The physiological signals obtained from
these sensors were amplied and ltered. After all preprocessing, the
signals were input to a microcontroller for nal processing. The basic
P. Kumari et al.
Fig. 8. Human Emotion Recognition System (Quazi, 2012).
emotions considered in this work were happy, sad, angry, relaxed. The
concept of smart skins has been proposed in some recent works (Cook
et al., 2013). Smart skins are intelligent cognitive skins that sense,
wirelessly communicate, and will be capable of modifying environ-
mental or physiological parameters using very simple passive RFID
technology sticker in the future. Smart skins are zero-power devices
meaning they rummage their own energy using ambient energy such as
electromagnetic, solar, mechanical, thermal, or RFID/Radar-based
inquisition techniques. The concept can also be extended as body-
wearable skins for continuous monitoring. One interesting application
of wearable sensors in sports and exercise is to use the data collected
from sensors for mapping real-world activities to the games, then,
developing the activity monitoring system in such a way that it can
produce an enjoyable game. Some authors have proposed the devel-
opment of a truly virtual mobile sports environment game tablet-
based exergame with wearable human body sensors (Mortazavi et al.,
2014). There was a time when sweat rate measurement for athletes was
possible only in laboratories. Now, it is possible to measure sweat rate
using wearable devices outside the laboratories also (Ermes et al.,
2008; Salvo et al., 2010; Strohrmann et al., 2012). It has been found
that exergames can be helpful to strengthen the children with autism
spectrum disorders physically and mentally. A novel wearable sensor
based Automatic Ingestion Monitoring (AIM) system has been reported
for unbiased monitoring of ingestive activity/behavior in normal life
(Fontana et al., 2014). It integrates three sensor modalities namely a
sensor for jaw motion, a sensor for hand gesture recognition, and an
accelerometer. These sensors interface to a smartphone through
wireless network. The development was aimed to correct known
ingestive behaviors causing weight gain with the help of a novel
behavioral modication tool. It will also help the patients with eating
disorders by studying their free-living food consumption. The possibi-
lity of wearable computers has been discussed which could share
thoughts and sensations (Bleicher, 2014). It would be like wearing a
computer on our arm. This will be personal and intimate and will be
based on technologies that attempt to colonize our whole body. It may
track our body movements, listen to our heartbeat and put our body on
line. A patent MD-based Brain Sentry represents a crucial concept of
wearable sensors which uses accelerometers to measure acceleration
during physical activity. The Brain Sentry helmet-mounted sensors give
an alert alarm when an athlete suers an unusual rapid and potentially
dangerous acceleration of the head. The alarm indicates possible injury
in the head of the player. The sensor is compact enough and weighs
only one ounce. It is waterproof and the batteries are long-lasting. It
can work for a whole year without charging, meaning no maintenance
by the person (Boyle, 2014). Mindful wearables have been designed
for improving physical health as well as mental well-being. Meditation
apps together with conventional EEG are being designed with the aim
to help in concentration build-up and self-regulation skills (Fernandez,
2015). To recognize the facial expression of human being, Step-wise
Linear Discriminant Analysis (LDA) and Hidden Conditional Random
Fields (HCRF) had been used and validated oine. However, in real-
world environment, its performance is not investigated till now. To
solve these problems, a robust, real-time subdivision technique may be
needed. Furthermore, in an actual environment, the facial frames may
303
Biosensors and Bioelectronics 90 (2017) 298307
Fig. 9. Evolution of EEG from conventional system to In-Ear-EEG.
have dierent angles and side views. Therefore, there is a need for
further research for improving the recognition rate of facial expressions
with various angles and litter of face (Siddiqi et al., 2015). Human body
movement can be recognised and separated with the use of Principle
Component Analysis (PCA) and Micro-Doppler (m-D) eect but, its
eectiveness is still to be veried for real-time application (Shi et al.,
2016; Mukhopadhyay, 2015). Human activity recognition has been in
some way, individualized. It means, the most of the systems can be
used only for a single user at a time. Although social networks give
information eective to recognize collective activities (Lane et al.,
2011) which can be considered one step further. A exible force sensor
was developed to assess the feasibility of detection of improper sitting
posture. The sensor was based on piezo-resistive capacitive lm (Lee
and Shin, 2016). The study can be further extended in order to increase
the resolution with the increase in number of sensors and the area of
measurement. The sensor platform can be optimized by determining
the size and position of the sensor. A wearable EEG was introduced
based on the In-Ear-EEG recording concept.
It was a generic device and can be used to record all the responses
that may be acquired from standard EEG (Goverdovsky et al., 2016).
Fig. 9 shows the evaluation of EEG system from conventional to In-
Ear-EEG.
Recently, a multi-tasking wearable device is developed which
monitors the health of person's heart as well as assess the position of
the person (Mao et al., 2008). To monitor dental cavities and other
dental plaque, a health monitoring apparatus is designed. It is an oral
monitoring system (Nanjundappa et al., 2016). A data fusion algorithm
based fall detection wearable system is developed for elderly persons.
The developed device uses MEMS, Kalman lter, Code Division
Multiple Access (CDMA) network to achieve ecient detection
(Wang and Qin, 2016). In very recent years, people are working on
resolving all these above mentioned major challenges for wearable
computing such as power line interference removal, selection of best
wearing location (Chimeno and Palls-areny, 2000). For continuous
monitoring of physiological signals with wearable devices,
Power Line interference (PLI) removal is of important considera-
tion. Keeping this in mind, a work has been presented for evaluation of
dierent techniques of PLI ltering to acquire better quality in
heterogeneous physiological signals (Tomasini et al., 2016).
A number of neurological disorders can be diagnosed with bio-
potential signals EEG, EOG, and EMG. For ECG recording, we have
many wireless systems as shown in Table 1. However, for bio-potential
signals such as EEG, EOG, and EMG recording, most of the platforms
provide wired system. Only few manufacturers such as Compumedics
provide Bluetooth system and Siesta's Ethernet radio link. A list is
shown in Table 2.
Sometimes, wearing these add-on devices in public places may be
uneasy or uncomfortable to the user (Knight and Baber, 2005). In a
P. Kumari et al.
Table 1
Commercial ECG recording platforms (Varadan et al., 2016).
Manufactures Product Name
Phillips Page writer TC50 ECG
GE Healthcare MARS Ambulatory ECG System (SEER 12, SEER light)
Imec Secure Digital Input Output
AliveCor AliveCor
Phillips EASI (Philips DigiTrak XT)
study, some common wearing locations were investigated. According to
statistical results, a noticeable alteration in elderly people's attitudes is
observed towards these add-on devices attached to dierent locations
on their body. The wrist was the most acceptable location for wearable
device by the users (Fang and Chang, 2016).
Many medical research also points to the importance of doctor-
patient interaction (Chaudhry et al., 2006; Cloudy Forecast:
Predictions for Big Data in Large and Small Medical Practices,
2013; Niksch et al., 2014). Whether the use of wearables by patients
might aect the doctor-patient interaction or whether this would
inuence patient behaviors too and therefore health results are
important questions (Loos and Davidson, 2016).
Neurological disorders like sleep disorder, drowsiness can also be
monitored by wearable wireless systems. According to National
Institute of Health (NIH), millions of Americans suer from sleep
disorders (Shen et al., 2006). For diagnosing sleep disorders, dierent
approaches are used such as polysomnography (PSG) and home sleep
test (HST). PSG is considered as a standard approach for monitoring
sleep disorders or sleep patterns. But this approach has many
disadvantages like: cost involved is very high, the patient has to stay
in the laboratory for the whole night which may cause discomfort or
inconvenience to the patient and that may also aect the test. Also the
waiting time for the test is very long due to unavailability of sucient
capacity of sleep laboratories (Flemons et al., 2004). The HST is
performed at home. So, it has overcome the problem of inconvenience
of patient, long waiting time and cost involved. But it has lack of real
time (RT) monitoring of sleep patterns and less physiological informa-
tion. The real time monitoring of signals from remote locations can be
achieved by the concept of combining two wireless communication
standards, on is for building WPAN (Wireless Personal Area Network)/
WBAN (Wireless Body Area Network) and other is for WAN (Wide
Area Network). Wireless wearable biosensors can also be used to
monitor drowsiness which is very useful in improving safety. People,
like security guards, night shift workers, drivers who need to stay
awake during night need drowsiness monitoring systems. According to
National Highway Trac safety Administration (NHTSA), approxi-
mately 25% of road accidents involve some form of drowsiness,
inattention, sleep or fatigue (Ranney et al., n.d.; The Role of Driver
Fatigue in Commercial Road Transport Crashes, 2001). Psychomotor
Vigilance Test (PVT) is one of the tests used to assess the quality of
sleep (Kamen, 2004). In a study, it has been found that continuous eye
gaze triggers the Autonomic nervous system (ANS) which causes
Table 2
Commercial EEG, EOG and EMG recording platforms (Varadan et al., 2016).
Manufactures Product Name Number of channels
Philips/Respironics Alice PDx 21channelwith optional ECG and EEG
Embla EmblettaX 100 12- channel with X100 proxy
Compumedics SomtePSG 16- channel
Compumedics Siesta 32- amplified channel
Cleveland Medical Sleep Scout 9-channel
ResMed ApneaLink Plus 4-channel
CareFusion Nox-T3 14- channel
Biosensors and Bioelectronics 90 (2017) 298307
No. of channels Storage Wire less
12 channels of ECG USB memory stick (upto 16 GB) No
312 channels of ECG 1 GB internal and optical DVD storage No
1channel ECG 16 GB Yes
1 channel ECG 16 GB Yes
4 channels ECG 256512 MB No
hormones release in blood stream and alters blood pressure and heart
rate. This eye gaze-induced stress level can be evaluated by monitoring
bio potential signals ECG and EOG (Aggarwal et al., 2014). To help the
elderly or disabled persons, a real-time EOG based alarm system has
been designed. Further, the work can be extended to various other
control applications for needy people (Bobade and Khirwadkar, 2016).
EOG technique is used to measure the resting potential of retina (Mala
and Latha, 2014). Wide applications of EOG also include activity
recognition (Bulling et al., 2011), virtual key board (Usakli and Gurkan,
2010), controlling of expert multitask gadget (Gandhi et al., 2010) and
many more.
Now a days, monitoring of human activity through fusion of a
number of bio potential signals are very much popular. A multimodal
fusion of EOG and EEG information has been shown as a robust
approach for recognition of reading activity in day-to-day life (Bulling
et al., 2012). For driving fatigue detection, a fusion of EEG signal and
forehead EOG signal (Huo et al., 2016) has been used which shows
better performance than solely using EEG (Wang et al., 2016) and
forehead EOG signal (Zhang et al., 2015). A wearable wireless brain-
computer interface system was developed for hybrid control of a robot
based on EEG and EOG signals (Oh et al., 2012). In this system, speed
and stopping of the robot were controlled by brain signals (EEG) and
right and left movements were controlled by EOG signals. A human-
machine interface system was developed which was based on wearable
sensor and only one channel EOG. The usability of the system is much
higher as it can enhance the comfort and wearability of the system
(Guo et al., 2016). In this work, individually optimized threshold for
each subject improved the classication performance, but the obtained
classication accuracy (84%) and Information Transfer Rate (ITR)
(13 bits/min) was not sucient to use this system for real-world
applications. Electrooculogram (EOG), electromyogram (EMG), and
glossokinetic potential (GKP) based multimodal interface was used for
human-computer interface to improve the information capacity. These
kinds of multimodal interface approach may improve the information
capacity as well as its robustness and usability. The author (Nam et al.,
2014) himself has suggested that the performance and the capacity of
the developed system can further be enhanced if it is integrated with
other modalities such as EEG. A robot control system was developed
using two modalities Electrooculography and Electromyogram. EOG is
using for moving the robot joint angles and EMG is using for object
grasping. However, it was found that the EOG and EMG signals are
hard to discriminate for the robot control system (Sasaki et al., 2015).
Storage Wireless
1 GB SD card No
128 MB internal memory No
2 GB Compact Flash Bluetooth
Compact Flash Siesta's Ethernet radio link
SD card 2.42.484 GHz
15 MB internal memory No
1 GB SD card No
304
P. Kumari et al.
The combination of EOG interface with RFID technology in shared
control architecture was used in an assistive robot. With this approach,
interaction with objects can be possible without using any kind of
motor related movement and the number of commands were increased
that can be generated by the EOG interface alone (Iez et al., 2012).
6. Conclusions
It can be concluded from the above discussion that human activity
recognition is an emerging area of research which is moving towards
the development of an intelligent and smart Healthcare Platform
Integrated in small devices for providing comfort to the patients as
well as elderly persons, for their well-being and independent living. A
signicant commercial development and progress have been already
reported. Still, we have many challenges and unsolved issues. Presently
available wearable sensors have solved the purpose to some extent.
However, they are also not very much cost eective and ecient from
all the aspects which have been discussed in this paper. These unsolved
issues motivate the development of new wearables having some new
techniques which can be applicable in health-care systems or HAR
systems and solve the purpose eciently for more realistic environ-
ments. It is expected from previous studies that many more user-
acceptable, high-performance, low cost wearable devices will be avail-
able for recognition of variety of activities. Surveys also predict that the
interest, and consequently, the use of wearable devices will be
increased in near future. At the same time, the cost of wearable devices
is expected to fall due to its wide applications in real-time. As much as
technology will mature, novelty in medical science will be increased,
further increase in integration of medical sensors and electronics
instruments will be witnessed and consequently, wearables will be
further smaller. It will support home based physiological data collec-
tion, preventive healthcare programs and also facilitate remote care.
The future of medical sciences is likely to be packed in the huge
prospective of the world's smallest sensors and wearable devices
where technology shorten to lengthen, and a generation lifted on
gadgets waste away. To provide a better alternative to the healthcare
system, development of comfortable, low cost wearable devices are
needed that can measure and continuously record the electrical activity
of human body as well as can transmit the obtained data wirelessly to a
computer, where it will be displayed in real time. If activity samples of a
signicant number of users from a predened area like a city, or a state
could be gathered as samples, the information obtained from
these samples could be used to estimate health conditions, early
diagnosis of some disorders, exercise habits. Hence further, this type
of participatory human-centered application would be free from
economic-incentive-based method. The users will participate willingly
in the system since they would be receiving the suggestions, informa-
tion related to their health conditions and exercises to improve their
physiological performance. Such gathered data from a sucient
number of users can be used to train machine learning classication
algorithms or soft computing techniques to enhance the overall
accuracy and consequently, to improve health diagnosis. EOG signal
has wide applications in clinic as well as in engineering such as
diagnosis of eye injuries, eye diseases, eye-controlled engineering
devices viz. cursor mouse. In spite of that, the use of an EOG interface
is still not enough capable of performing tasks involving complex
control commands to guide an external device optimally. One solution
for these limitations is shared control approach. E.g. combining EOG
and Radio-frequency identication (RFID) technology. Shared control
has a wide range of applications. In robotics, it is used with multi-agent
systems on cooperative tasks. Shared control is also useful for disabled
people, e.g., to aid blind people and to improve mobility. In future
research, one can have the opportunity to work for minimization of
positions of EOG electrodes and design of real time wearable devices
based on EOG signal. To increase the performance of recognition, more
multimodal fusion of physiological signals is sought, since a single
305
Biosensors and Bioelectronics 90 (2017) 298307
physiological signal cannot have all the information required for a
particular task.
7. Scope of future work
There are a lot of works centered in solving accessibility issues
using rehabilitation robotics or developing alternative communication
methods with the environment. Humanmachine interfaces are used
to interact with external devices aimed at helping handicapped people
or elderly persons. Ocular movements can be used to generate
commands to control external devices as in progressive neurodegen-
erative diseases, the brain and spinal cord nerve cells are aected but
the eye movements remain intact. Since most completely paralyzed
people can still move their eyes, EOG is a signal that can be practically
used for controlling the prosthetic devices. The advantage of EOG in
terms of accuracy and complexity is quite important so it is one of the
most commonly used methods to detect eye movement. The combina-
tion of an EOG interface with Radio-frequency identication (RFID)
technology in shared control architecture is used in recent research
works. However, a single EMG, EEG or EOG-based system can only
manage one certain kind of task. It is rather dicult to have a
universally robust system applicable to dierent situations.
Therefore, multimodal system will be a more eective approach.
Shared control with multimodal interface can be a better approach
for generating complex control commands to optimally guide an
external device or application. Wearable systems can be designed to
control the dierent kind of objects based on multimodal interface and
shared control architecture where multimodal interface can take
decisions about the movement of a robot for a particular application
and the interaction with objects can be assisted by RFID technology.
More modalities can be added in future to further enhance the
controlling commands and make the product more feasible and usable.
References
Aggarwal, Y., Singh, N., Ghosh, S., Sinha, R.K., 2014. Eye gaze-induced mental stress
alters the heart rate variability analysis. J. Clin. Eng. 39, 7989. http://dx.doi.org/
10.1097/JCE.0000000000000023.
Alert Devices for the Elderly [WWW Document], 2016. Med. Alert Syst. Rev. Site. URL
http://medicalalertsystemshq.com/blog/medical?alert?devices?for?the?elderly.
html.
Bagade, P., Banerjee, A., Gupta, S.K.S., 2014. Optimal design for symbiotic wearable
wireless sensors. In: Proceedings of the 2014 11th International Conference on
Wearable and Implantable Body Sensor Networks, IEEE Computer Society. pp. 132
137.
Bleicher, A., 2014. Beyond words: Wearable computers will let us share thoughts and
sensations. IEEE Spectr., 6671.
Bobade, J.R., Khirwadkar, M.D., 2016. Design and implementation of electrooculogram
based alarm system for disabled. Int. J. Adv. Res. Comput. Eng. Technol. 5,
872874.
Bonato, P., 2010. Wearable sensors and systems. IEEE Eng. Med. Biol. Mag., 2536.
http://dx.doi.org/10.1109/MEMB.2010.936554.
Boyle, J.A., 2014. Brain Sentry Licenses Breakthrough Wearable Sensor Technology from
Johns Hopkins University.
Bulling, A., Ward, J., Gellersen, H., Troster, G., 2011. Eye movement analysis for activity
recognition using electrooculography. IEEE Trans. Pattern Anal. Mach. Intell. 33,
741753.
Bulling, A., Ward, J.A., Gellersen, H., 2012. Multimodal recognition of reading activity in
transit using body-worn sensors. ACM Trans. Appl. Percept. 9, 2:12:21. http://
dx.doi.org/10.1145/2134203.2134205.
Chaudhry, B., Wang, J., Wu, S., Maglione, M., Mojica, W., Roth, E., Morton, S.C.,
Shekelle, paul G., 2006. Systematic review: impact of health information technology
on quality, eciency, and costs of medical care. Ann. Intern. Med. 144, 742752.
Chen, B., Patel, S., Buckley, T., Rednic, R., Mcclure, D.J., Shih, L., Tarsy, D., Welsh, M.,
Bonato, P., 2011. A web-based system for home monitoring of patients with
Parkinson s disease using wearable sensors. IEEE Trans. Biomed. Eng. 58, 831836.
Chimeno, M.F., Palls-areny, R., 2000. A comprehensive model for power line
interference in biopotential measurements. IEEE Trans. Instrum. Meas. 49,
535540.
Choujaa, D., Dulay, N., 2008. TRAcME: temporal activity recognition using mobile phone
data. In: Proceedings of the IEEE/IFIP International Conference on Embedded
Ubiquitous Computing. pp. 119126. http://dx.doi.org/10.1109/EUC.2008.33.
Cloudy Forecast : Predictions for Big Data in Large and Small Medical Practices [WWW
Document], 2013. Cern. Ambul.
Cook, B.S., Le, T., Palacios, S., Traille, A., Tentzeris, M.M., 2013. Only skin deep. IEEE
P. Kumari et al.
Macrow. Mag., 103114.
Cunningham, P., Delany, S.J., 2007. k-Nearest Neighbour Classiers. Dublin.
Englebienne, T.L.M.V.K.G., Krose, B.J.A., 2010. An activity monitoring system for elderly
care using generative and discriminative models. J. Pers. Ubiquitous Comput. 14,
489498. http://dx.doi.org/10.1007/s00779-009-0277-9.
Ermes, M., Parkka, J., Mantyjarvi, J., Korhonen, I., 2008. Detection of daily activities and
sports with wearable sensors in controlled and uncontrolled conditions. IEEE Trans.
Inf. Technol. Biomed. 12, 2026.
Fang, Y.-M., Chang, C.-C., 2016. Users' psychological perception and perceived
readability of wearable devices for elderly people. Behav. Inf. Technol., 18. http://
dx.doi.org/10.1080/0144929X.2015.1114145.
Fernandez, A., 2015. Mindful wearables [WWW Document]. Sharp Brains. URL
http://sharpbrains.com/blog/2015/11/10/10-neurotechnologies-about-to-
transform-brain-enhancement-and-brain-health/.
Flemons, W.W., Douglas, N.J., Kuna, S.T., Rodenstein, D.O., Wheatley, J., 2004. Access
to diagnosis and treatment of patients with suspected sleep apnea. Am. J. Respir.
Crit. Care Med. 169, 668672. http://dx.doi.org/10.1164/rccm.200308-1124PP.
Fontana, J.M., Farooq, M., Sazonov, E., 2014. Automatic ingestion monitor: a novel
wearable device for monitoring of ingestive behavior. IEEE Trans. Biomed. Eng. 61,
17721779.
Gandhi, T., Trikha, M., Santhosh, J., Anand, S., 2010. Development of an expert
multitask gadget controlled by voluntary eye movements. Expert Syst. Appl. 37,
42044211. http://dx.doi.org/10.1016/j.eswa.2009.11.082.
Goverdovsky, V., Looney, D., Kidmose, P., Mandic, D.P., 2016. In-ear EEG from
viscoelastic generic earpieces: robust and unobtrusive 24/7 monitoring. IEEE Sens.
J. 16, 271277.
Gunn, S.R., 1998. Support Vector Machines for Classication and Regression.
Guo, X., Pei, W., Wang, Y., Chen, Y., Zhang, H., Wu, X., Yang, X., Chen, H., Liu, Y., Liu,
R., 2016. A human-machine interface based on single channel EOG and patchable
sensor. Biomed. Signal Process. Control 30, 98105. http://dx.doi.org/10.1016/
j.bspc.2016.06.018.
Hayward, J., Chansin, D.G., 2016. Wearable Sensors 2016-2026: Market Forecasts,
Technologies, Players [WWW Document]. URL http://www.idtechex.com/
research/reports/wearable-sensors-2016-2026-market-forecasts-technologies-
players-000470.asp.
Hayward, J., Das, R., Holland, G., 2015. Wearable Technology 20152025 : Technologies
, Markets, Forecasts [WWW Document]. URL http://www.idtechex.com/research/
reports/wearable?technology?2015?2025?technologies?markets?forecasts?000427.
asp.
Hopeld, J.J., 1988. Articial neural networks. IEEE Circuits Devices Mag., 310.
Huo, X., Zheng, W., Lu, B., 2016. Driving fatigue detection with fusion of EEG and
forehead EOG. In: Proceeding of International Joint Conference on Neural Networks
(IJCNN-16).
Iez, E., beda, A., Azorn, J.M., Perez-vidal, C., 2012. Assistive robot application based
on an RFID control architecture and a wireless EOG interface. Rob. Auton. Syst. 60,
10691077. http://dx.doi.org/10.1016/j.robot.2012.05.006.
Iglesias, J., Cano, J., Bernardos, A.M., Casar, J.R., 2011. A ubiquitous activity-monitor to
prevent sedentariness. IEEE Conf. Pervasive Comput. Commun., 319321.
Jatob, L.C., Gromann, U., Kunze, C., Ottenbacher, J., Stork, W., System, A.M., 2008.
Context-aware mobile health monitoring: evaluation of dierent pattern recognition
methods for classication of physical activity. Annu. Int. IEEE EMBS Conf.,
52505253.
Juha, P., Ermes, M., Korpipaa, P., Mantyjarvi, J., Peltola, J., 2006. Activity classication
using realistic data From Wearable sensors. IEEE Trans. Inf. Technol. Biomed. 10,
119128.
Kamen, G., 2004. Electromyographic kinesiology. In: Robertson, L.D., Tocco, A.N. (Eds.),
Research Methods in Biomechanics. Human Kinetics Publication, Champaign, IL,
171181.
Knight, J.F., Baber, C., 2005. A tool to assess the comfort of wearable computers. Hum.
Factors Ergon. Soc. 47, 7791. http://dx.doi.org/10.1518/0018720053653875.
Lane, N.D., Xu, Y., Campbell, A.T., Choudhury, T., Eisenman, S.B., 2011. Exploiting
social networks for large-scale human behavior modeling. IEEE Pervasive Comput.
10, 4553.
Lara, D., Labrador, M.A., 2013. A survey on human activity recognition using wearable
sensors. IEEE Commun. Surv. Tutor. 15, 11921209.
Lee, B.W., Shin, H., 2016. Feasibility study of sitting posture monitoring based on
piezoresistive conductive lm-based exible force sensor. IEEE Sens. J. 16, 1516.
Lin, C., Yang, Y.C., Wang, J., Yang, Y., 2012. A wearable sensor module with a neural-
network- based activity classication algorithm for daily energy expenditure
estimation. IEEE Trans. Inf. Technol. Biomed. 16, 991998.
Lin, Y., 2009. Decision Making in assistive environments using multimodal observations.
In: Proceedings of the Petra '09 ACM. pp. 18.
Liu, G., Huang, B., Wang, L., 2011. A wearable respiratory biofeedback system based on
generalized body sensor network. Telemed. J. E. Health, 348357. http://
dx.doi.org/10.1089/tmj.2010.0182.
Loos, J.R., Davidson, E.J., 2016. Wearable health monitors and physician-patient
communication: the physicians perspective. In: Proceeding of the 2016 49th Hawaii
International Conference on System Sciences (HICSS)HICSS '16. pp. 33893399.
http://dx.doi.org/10.1109/EUC.2008.3310.1109/HICSS.2016.422.
Mala, S., Latha, K., 2014. Feature selection in classication of eye movements using
electrooculography for activity recognition. Comput. Math. Methods Med. 2014, 19
. http://dx.doi.org/10.1155/2014/713818.
Malhi, K., 2010. Wireless Sensors Network Based Physiological Parameters Monitoring
System. Massey University, New Zealand.
Mao, M., Chuo, Y., Kaminska, B., 2008. Multi-functional wearable device for heart health
and position assessment. NSTI Nanotech. 2, 615618.
306
Biosensors and Bioelectronics 90 (2017) 298307
Mariani, B., Jimenez, M.C., Vingerhoets, F.J.G., Aminian, K., 2013. On-shoe wearable
sensors for gait and turning assessment of patients With Parkinsons disease. IEEE
Trans. Biomed. Eng. 60, 155158.
Marquardt, D., 1963. an algorithm for least-squares estimation of nonlinear
paramenters. J. Soc. Ind. Appl. Math. 11, 431441.
Mobile phones to come with panic button from 2017, 2016. Econ. Times Bur. pp. 25.
Mortazavi, B., Nyamathi, S., Lee, S.I., Wilkerson, T., Ghasemzadeh, H., Sarrafzadeh, M.,
2014. Near-realistic mobile exergames with wireless wearable sensors. IEEE J.
Biomed. Heal. Inform. 18, 449456.
Mukhopadhyay, S.C., 2015. Wearable sensors for human activity monitoring: a review.
IEEE Sens. J. 15, 13211330.
Nam, Y., Koo, B., Cichocki, A., Choi, S., 2014. GOM-face: GKP, EOG, and EMG-based
multimodal interface with application to humanoid robot control. IEEE Trans.
Biomed. Eng. 61, 453462. http://dx.doi.org/10.1109/TBME.2013.2280900.
Nanjundappa, R., Franklin, R., Dharmapurikar, V.A., Singh, S., 2016. Oral health
Monitoring Method and apparatus and Electronic Device Using the Same.
Nemati, E., Deen, M.J., Mondal, T., 2012. A wireless wearable ECG sensor for long-term
applications. IEEE Commun. Mag., 3643.
Niksch, A.L., Rothman, B.S., Davidson, S.J., 2014. The value of remote patient
monitoring (RPM) physicians' perspectives. Healthc. Inf. Manag. Syst. Soc., 15.
Oh, S., Kumar, P.S., Kwon, H., Varadan, V.K., 2012. Wireless brain-machine interface
using EEG and EOG: Brain wave classication and robot control. In: SPIE
Proceedings on Nanosensors, Biosensors, and Info-Tech Sensors and Systems. pp.
18. http://dx.doi.org/10.1117/12.918159
Patel, S., Lorincz, K., Hughes, R., Huggins, N., Growdon, J., Standaert, D., Akay, M., Dy,
J., Welsh, M., Bonato, P., Member, S., 2009. Monitoring motor uctuations in
patients with Parkinson s disease using wearable sensors. IEEE Trans. Inf. Technol.
Biomed. 13, 864873.
Patel, S., Park, H., Bonato, P., Chan, L., Rodgers, M., 2012. A review of wearable sensors
and systems with application in rehabilitation. J. Neuroeng. Rehabil. 9, 21. http://
dx.doi.org/10.1186/1743-0003-9-21.
Peixoto, P., Batista, J., Araujo, H.J., 2001. Real-time human activity monitoring
exploring multiple vision sensors. Rob. Auton. Syst. 35, 221228. http://dx.doi.org/
10.1016/S0921-8890(01)00117-8.
Quazi, M.T., 2012. Human Emotion Recognition Using Smart Sensors. Massey
university, New Zealand.
Ramasamy, V., Gowda, C., Noopuran, S., 2014. The basics of designing wearable
electronics with microcontrollers [WWW Document]. Cypress Semicond. URL
http://www.embedded.com/print/4431259
Ranney, T.A., Garrott, W.R., Goodman, M.J., n.d. NHTSA Driver Distraction Research:
Past, Present and Future. 2001 pp. 18, http://www-nrd.nhtsa.dot.gov/pdf/esv/
esv17/proceed/00177.pdf
Rodgers, M.M., Pai, V.M., Conroy, R.S., 2015. Recent advances in wearable sensors for
health monitoring. IEEE Sens. J. 15, 31193126.
Salvo, P., Francesco, F., Di, Costanzo, D., Ferrari, C., Trivella, M.G., Rossi, D. De, 2010. A
wearable sensor for measuring sweat rate. IEEE Sens. J. 10, 15571558.
Sarkar, J., Vinh, L.T., Lee, V.Y., 2011. GPARS: a general-purpose activity recognition
system. Appl. Intell. 35, 242259. http://dx.doi.org/10.1007/s10489-010-0217-4.
Sasaki, M., SAuhaimi, M.S.B.A., Matsushita, K., Ito, S., Rusydi, M.I., 2015. Robot control
system based on electrooculography and electromyogram. J. Comput. Commun. 3,
113120.
Seoane, F., Mohino-herranz, I., Ferreira, J., Alvarez, L., Llerena, C., Gil-pita, R., 2014.
Wearable biomedical measurement systems for assessment of mental stress of
combatants in real time. Sensors 40, 71207141. http://dx.doi.org/10.3390/
s140407120.
Shen, J., Barbera, J., Shapiro, C.M., 2006. Distinguishing sleepiness and fatigue: focus on
denition and measurement. Sleep Med. Rev. 10, 6376. http://dx.doi.org/10.1016/
j.smrv.2005.05.004.
Shi, X., Zhou, F., Tao, M., Zhang, Z., 2016. Human movements separation based on
principle component analysis. IEEE Sens. J. 16, 20172027.
Siddiqi, M.H., Ali, R., Khan, A.M., Park, Y., Lee, S., 2015. Human facial expression
recognition using stepwise linear discriminant analysis and hidden conditional
random elds. IEEE Trans. Image Process. 24, 13861398.
Strohrmann, C., Harms, H., Kappeler-setz, C., Tr, G., 2012. Monitoring kinematic
changes with fatigue in running using body-worn sensors. IEEE Trans. Inf. Technol.
Biomed. 16, 983990.
The Role of Driver Fatigue in Commercial Road Transport Crashes, 2001.
Toh, W.Y., Tan, Y.K., Koh, W.S., Siek, L., 2014. Autonomous wearable sensor nodes with
exible energy harvesting. IEEE Sens. J. 14, 22992306.
Tolstikov, A., Hong, X., Biswas, J., Nugent, C., Chen, L., Parente, G., 2011. Comparison of
fusion methods based on DST and DBN in human activity recognition. J. Control
Theory Appl. 9, 1827. http://dx.doi.org/10.1007/s11768-011-0260-7.
Tomasini, M., Benatti, S., Milosevic, B., Farella, E., Benini, L., 2016. power line
interference removal for high-quality continuous biosignal monitoring with low-
power wearable devices. IEEE Sens. J. 16, 38873895.
Ugulino, W., Cardador, D., Vega, K., Velloso, E., Milidiu, R., Fuks, H., 2012. Wearable
computing: accelerometers' data classication of body postures and movements.
Adv. Artif. Intell. 7589, 5261. http://dx.doi.org/10.1007/978-3-642-34459-6.
Usakli, A.B., Gurkan, S., 2010. Design of a novel ecient human computer interface:
an electrooculagram based virtual keyboard. IEEE Trans. Instrum. Meas. 59,
20992108.
Uslu, G., Dursunoglu, H.I., Altun, O., Baydere, S., 2013. Human activity monitoring with
wearable sensors and hybrid classiers. Int. J. Comput. Inf. Syst. Ind. 5, 345353.
Varadan, V.K., Rai, P., Oh, S.C., Kumar, P.S., 2016. Wearable technology and mobile
platform for human health monitoring. Forum Electromagn. Res. Methods Appl.
Technol. 16, 138.
P. Kumari et al.
Velloso, E., Bulling, A., Gellersen, H., Ugulino, W., 2013. Qualitative activity recognition
of weight lifting exercises. In: Proceeding of the International Conference in
Cooperation with SIGCHI (Augmented Human13).
Versel, N., 2012. Panic buttons for seniors must go [WWW Document].
MobiHealthNews. URL http://mobihealthnews.com/19397/panic?Buttons?For?
Seniors?Must?Go.
Wang, L., Zheng, W., Ma, H., Lu, B., 2016. Measuring sleep quality from EEG with
machine learning approaches. In: Proceedings of the International Joint Conference
on Neural Networks (IJCNN-16).
307
Biosensors and Bioelectronics 90 (2017) 298307
Wang, X., Qin, H., 2016. Implementation of fall detection system based on data fusion
technology. Int. J. u- e- Serv. Sci. Technol. 9, 18. http://dx.doi.org/10.14257/
ijunesst.2016.9.4.01.
Yang, J., Lee, J., Choi, J., 2011. Activity recognition based on RFID object usage for
smart mobile devices. J. Comput. Sci. Technol. 26, 239246. http://dx.doi.org/
10.1007/s11390-011-1126-7.
Zhang, Y., Gao, X., Zhu, J., Zheng, W., Lu, B., 2015. A Novel Approach to Driving Fatigue
Detection Using Forehead EOG. In: Proceedings of the International IEEE/EMBS
Conference on Neural Engineering. pp. 707710.